In recent years, light-sheet microscopy, or Single Plane Illumination Microscopy (SPIM), has reemerged as a valuable imaging technique, offering considerable advantages over conventional microscopy methods in developmental studies (Chardes et al., 2014; Huisken and Stainier, 2007; Keller, 2013; Keller et al., 2008, 2011; Mertz and Kim, 2010; Santi et al., 2009; Tomer et al., 2012; Truong et al., 2011; Weber and Huisken, 2011). SPIM employs side-on sample illumination using a micrometer-thick laser sheet to generate optical sections; fluorescent images are collected by a camera with its optical axis orthogonal to the illumination plane. Different regions in the focal plane are simultaneously illuminated, reducing the overall exposure time of samples, allowing faster throughput times and longer-term imaging protocols with minimal sample damage.

Acquisition of three-dimensional tissue data has been extensively done using confocal or two-photon microscopy by us and others during zebrafish embryogenesis (Kwak et al., 2013). These methods however are limited in speed and depth of imaging: only relatively small tissue samples can be scanned, meaning that whole brain samples, even those of the embryonic zebrafish cannot be imaged in their entirety. Photobleaching and phototoxicity also affect long-term in vivo imaging by confocal microscopy.

Light-sheet microscopy methodologies however can provide single-cell, or near single-cell resolution detail of a large proportion of the embryonic or larval zebrafish brain volume. Acquisition speed using SPIM is also 20?50 fold faster than point-scanning imaging without compromising the signal-to-noise ratio, resulting in reduced photobleaching and phototoxicity (Keller et al., 2011; Panier et al., 2013). Recently, functional light-sheet microscopy technology has been developed, allowing whole-brain mapping of neural activity in zebrafish embryos (Ahrens et al., 2013; Panier et al., 2013). The extended field of view provided by light-sheet microscopy allows the direct identification of related groups of cells spanning whole brain regions, helping establish interactive relationships between cells and tissues likely to be important during animal development.

Analysis of upstream genomic elements in neuronal and vascular promoters has led to the establishment of transgenic zebrafish lines with fluorescent neurons, blood cells or blood vessels (Isogai et al., 2001; Park et al., 2000; Traver et al., 2003). These resources are powerful tools, permitting the three dimensional mapping of distinct tissues in the whole brain and body. The vascular anatomy of the developing zebrafish has been previously described in detail using confocal imaging (Isogai et al., 2001).

Here, we used live SPIM imaging and transgenic lines to visualize and map in 3D the blood vessels and differentiating neurons of the entire zebrafish brain during development. We also imaged circulating blood cells within blood vessels of the whole anterior embryonic body in 3D. Our results reveal details of the spatial relationship between the developing nervous system and blood vessels, as well as providing a direct visualization of blood flow through the entire cranial vasculature. Such data cannot be obtained by traditional point scanning microscopy.

Zebrafish husbandry

Adult zebrafish (Danio rerio) were maintained at 27.5°C in a 14/10 h light/dark cycle. Embryos were raised in E3 medium (5 mM NaCl, 0.33 mM MgSO₄, 0.33 mM CaCl₂, and 0.17 mM KCl) at 28.5°C and staged by hours post fertilization according to Kimmel et al. (1995). Strains used in this study were wild type AB, Tg(elavl3:DsRed)(Paquet et al., 2009), Tg(fli1a:EGFP) (Lawson et al., 2002) and Tg(gata1: DsRed)(Traver et al., 2003). All experiments were conducted according to protocols approved by the Institutional Animal Care.

Zebrafish embryo embedding for imaging

Zebrafish embryos were first anesthetized in E3 media containing 0.02% Tricaine (Sigma), then embedded and centered within glass capillaries (1.0 mm inner diameter, Zeiss) filled with low-melting point agarose (LMP, 2% in E3 medium, 0.02% Tricaine). After setting, an agarose cylinder containing the anterior portion of the embryo was extruded from the capillary by inserting the opposite end into a microfuge tube containing 2% LMP agarose. The imaging chamber was filled with E3 medium 0.02% Tricaine and was maintained at 23°C throughout the imaging period.

Light-sheet microscopy imaging

In all experiments a Lightsheet Z.1 fluorescence microscope with W Plan-Apochromat 20×/1.0 objective (Zeiss) was used to collect data. Fluorescence was excited with 488 nm and 561 nm lasers, and emission was detected with 500?545 nm and 575?615 nm band-pass filters.

Image processing

All SPIM data was saved in the LSM format and processed using ZEN software (Zeiss). Maximum intensity projections were generated from each z-stack file. Three-dimensional images were reconstructed using Imaris software (Bitplane, USA), in which each voxel was recognized by the power of each fluorescent signal.

Morphogenesis of the nervous system and vasculature is coordinated during embryogenesis; development of these two tissue types can occur independently, but key interactions may occur in particular regions at specific times which are necessary for proper development and function. To fully understand the activity-dependent interactions between these tissues, or to map neural circuits, precise 3D positional data across developmental time is required across the entire head or body. Here, we performed simultaneous live 3D imaging of cranial neurons and blood vessels in the entire embryonic brains of transgenic zebrafish using light-sheet fluorescence microscopy to maximize physical coverage of the brain.

Light-sheet microscopy of cranial neurons and blood vessels at 24 hpf

The transgenic line Tg(fli1a:EGFP) fluorescently labels the developing vasculature and pharyngeal arches. This was combined with Tg(elavl3:DsRed) animals, in which DsRed fluorescent protein is expressed under the control of the pan-neuronal elavl3/HuC promoter (Park et al., 2000). The resultant Tg (elavl3:DsRed)/Tg(fli1a:EGFP) embryos were live imaged in 3D at 24 hpf (Fig. 1). Using Imaris software, 3D images were generated from the entire stack of 1740 optical sections (0.488 μm thickness) (Fig. 1A, a), or from selected central optical sections to more clearly show internal structures (Fig. 1A, b?d). Samples are positioned in rotatable glass capillaries within the Lightsheet Z.1 microscope, allowing imaging from all sides. 3D reconstructions of the 24hpf embryonic head were generated illustrating dorsal, ventral and lateral views (Fig. 1B, a,c,e), each of these was further processed to create a 3D segmented model of fli1a:EGFP-positive tissues (Fig. 1B, b,d), or both fli1a:EGFP-positive and elavl3:DsRed-positive tissues (Fig. 1B, f) by surface rendering. An animated 3D reconstruction is presented in supplementary movie 1 (Supplementary Fig. S1), revealing the detailed internal structures of embryonic tissues in the whole head and anterior trunk.

At 24 hpf, the embryo is curled around the dorsal yolk surface; it therefore takes longer to completely image the entire embryo using longitudinal optical sections than older, straighter, embryos. In this experiment, light-sheet imaging to a depth of 850 μM covered the entire head and anterior trunk (1740 optical sections of 0.488 μm thickness). In contrast, confocal laser scanning (LSM780, Zeiss) of the same sample could only achieve a 3D image depth of 100?200 μM (data not shown).

Blood cells begin to circulate through the developing blood vessels at this stage. The arch aorta extends to the lateral dorsal aorta, then merges caudally with the dorsal aorta (Figs. 1A, c and 2B, s,d,f). The dorsal aorta runs caudally to the anal pore, while the posterior cardinal vein runs in the opposite direction from the anal pore to anterior head. Intersegmental veins branch dorsally, and the common cardinal veins develop bilaterally on the dorsal yolk surface.

3D SPIM imaging and reconstruction of cranial neurons and blood vessels at 42 hpf

SPIM imaging of 42 hpf embryos was also performed (Fig. 2). In this case, light sheet thickness was 1.176 μm for each section and again, selected central sections were reconstructed to more clearly visualize internal structures (Fig. 2A, a?e, 479 sections; Fig. 2B, a?r”, 420 sections). Imaris software was used to generate a solid surface model, illustrating deep structures inside the brain (Fig. 2A, e), and a 3D reconstruction of total raw data (Fig. 2A, d). Animated 3D reconstructions of the complete 42 hpf SPIM data (Supplementary Fig. S2A), the central SPIM data after removal of most dorsal and ventral sections (Supplementary Fig. S2B), and the same central SPIM data following segmentation (Supplementary Fig. S2C) is presented as supplementary movie files.

The detailed 3D morphology of the developing embryo is clearly evident from the most anterior veins and differentiating neurons to the spinal neurons and sprouting veins of the trunk (labeled in Fig. 2A,e).

3D networks of cranial neurons and vasculature at 3.2 dpf

Tg(elavl3:DsRed/ fli1a:EGFP) embryos were similarly imaged at 3.2 dpf from dorsal (Fig. 3A, a?c) and lateral (Fig. 3A, d?f) viewpoints. 503 dorsal sections of 1.0 ?m light-sheet thickness were assembled for 3D reconstruction. At this stage, the innermost structures most distant from the objective were slightly fuzzy, although the morphology of all developing neurons and blood vessels in the entire anterior embryo were identifiable (Fig. 3A, g; Supplementary movie Fig. S3A). Segmentation of the 3D image after removal of the most dorsal and ventral sections using Imaris surface rendering resulted in clearer visualization of internal structures (Fig. 3A, h; Supplementary movie Fig. S3B). Lateral imaging of the same 3.2 dpf Tg(fli1a: EGFP/elavl3:DsRed) embryo was also performed; the 3D reconstruction was generated from 591 sections of 1.0 ?m optical thickness, resulting in clearer presentation of the lateral structures (Fig. 3A, d?f).

Multi-view 3D imaging of blood cells and blood vessels at 35 hpf

Next, circulating blood cells and their associated vasculature were simultaneously visualized in live Tg(fli1a:EGFP/gata1: DsRed) embryos at 35 hpf (Fig. 4). In this transgenic animals, gata1:DsRed-positive blood cells circulate through fli1a:EGFP-positive blood vessels (Traver et al., 2003). Multi-view images of the whole brain, anterior trunk and the subintestinal vasculature around the yolk were recorded in 696 optical sections of 1.0 ?m thickness (Fig. 4). Light-sheet microscopy imaging of samples in glass tubes allows the collecting of optical sections from multiple directions, and therefore, the limitations of SPIM resolution in deeper sections of large specimens can be overcome. Using optical sections from dorsal view imaging, we generated a 3D reconstruction from raw data (Fig. 4, m; Supplementary Fig. S4A) and also constructed segmentation models using the surface function (Fig. 4, n; Supplementary Fig. S4B). This study demonstrates that light-sheet imaging allows the complete visualization of blood cell formation, local distribution and circulation through developing blood vessels in the entire zebrafish embryo during development.

This study is the first to visualize and comprehensively map the 3D anatomy of cranial neurons, vasculature tissues and blood cells in the entire embryonic brains at different stages of embryogenesis using Light-Sheet microscopy. SPIM overcomes many of the limitations of focal imaging, including confocal microscopy, generating live 3D images at high speed and spatio-temporal resolution. We achieved an imaging depth of 850 μM with SPIM, permitting complete imaging of the entire embryonic zebrafish head within 15 min. In comparison, confocal microscopy was only effective to a depth of 200 μM and complete sample imaging required hours of scanning time. Our results provide the foundation for the construction of a complete 4D developmental atlas of zebrafish embryogenesis; such a tool would be of particular benefit in providing reference for registering focal neural activity of specific brain regions across the entire brain, and for the study of detailed tissue anatomy and morphological process in normal and mutant fish.

The aim of next generation live imaging techniques should be to monitor the activities of complete neural circuits at high resolution over prolonged time periods. Multi-view light-sheet microscopy has so far been adopted to generate quantitative in vivo data from whole Drosophila and zebrafish embryos (Tomer et al., 2012). In mice, fast three-dimensional fluorescence imaging of activity in neural populations has been demonstrated by objective-coupled planar illumination microscopy (Holekamp et al., 2008). Exciting recent progress in optogenetics would be complemented by further refinement of available in vivo imaging methodologies, providing new ways of studying brain activity and associated behavioral changes in living animals. 3D maps such as ours will be invaluable in providing 3D anatomical landmarks for correlating optogenetic data with distinct neuronal circuits by rapid analysis of the entire nervous system and cranial vasculature in a single experiment; this offers real advantages over traditional landmarks which are based on anatomical features.

The volume of data created during light-sheet microscopy is far greater than that of standard confocal microscopy, often requiring terabytes of storage capacity per experiment, and thus computing power may currently be a limiting factor in the complete assembly and interpretation of light-sheet microscopy image files.